embodied-ai-news

Aggregates publicly available Embodied AI and Robotics news from curated sources (robotics media, arXiv, company blogs). Delivers structured briefings on humanoid robots, foundation models, hardware, deployments, and funding with direct links to original articles.

# Embodied AI News Briefing

> Aggregates the latest Embodied AI & Robotics news from curated sources and delivers concise summaries with direct links. Covers the full stack: algorithms, hardware, simulation, deployment, funding, policy, and the China ecosystem.

## When to Use This Skill

Activate this skill when the user:

- Asks for embodied AI news, robot news, or humanoid robot updates
- Requests a daily/weekly/monthly robotics briefing
- Mentions wanting to know what's happening in embodied AI / robotics
- Asks about specific companies: Tesla Optimus, Figure, Unitree, AGIBOT, Boston Dynamics, etc.
- Asks about specific technologies: VLA models, diffusion policy, sim-to-real, dexterous manipulation
- Wants a summary of recent robotics research papers
- Asks about robotics funding, deployments, or supply chain
- Asks about simulation platforms, benchmarks, or datasets
- Asks about robotics policy, safety standards, or export controls
- Requests a monthly trend report or competitive analysis
- Says: "给我今天的具身智能资讯" (Give me today's embodied AI news)
- Says: "机器人行业有什么新动态" (What's new in the robot industry)
- Says: "最近有什么人形机器人的消息" (Any recent humanoid robot news)
- Says: "这个月的具身智能趋势报告" (This month's embodied AI trend report)
- Says: "embodied AI updates", "robot learning news", "humanoid robot news"

### Trigger Keywords

**English**: `embodied AI`, `humanoid robot`, `robot news`, `robotics update`, `robot learning`, `VLA model`, `diffusion policy`, `dexterous manipulation`, `sim-to-real`, `robot deployment`, `robotics funding`, `Figure AI`, `Tesla Optimus`, `Unitree`, `AGIBOT`, `Boston Dynamics`, `1X`, `Physical Intelligence`, `Skild AI`, `robot hand`, `quadruped robot`, `Isaac Sim`, `world model robot`, `robot benchmark`, `robot safety`, `robot regulation`, `monthly robot report`

**Chinese**: `具身智能`, `人形机器人`, `机器人资讯`, `灵巧操作`, `仿真到真实`, `机器人部署`, `宇树`, `智元`, `优必选`, `银河通用`, `傅利叶`, `机器人融资`, `灵巧手`, `四足机器人`, `机器人大模型`, `机器人月报`, `机器人安全`, `机器人政策`

---

## Reference Files

This skill relies on 5 companion reference files. Always consult them during execution:

```
📁 references/
├── 📰 news_sources.md        — WHERE to find information (tiered source list)
├── 🔍 search_queries.md     — HOW to search (query templates & recipes)
├── 📝 output_templates.md   — WHAT format to output (6+ template variants)
├── 📊 taxonomy.md           — SHARED LANGUAGE (categories, keywords, company list)
└── 🧭 workflow.md           — WHEN and in what ORDER to execute (SOP for daily/weekly/monthly)
```

| File                  | When to Consult                                                                         |
| --------------------- | --------------------------------------------------------------------------------------- |
| `news_sources.md`     | Phase 1 — choosing which sites to fetch; selecting tier-appropriate sources             |
| `search_queries.md`   | Phase 1 — building search queries; selecting recipe by briefing type                    |
| `taxonomy.md`         | Phase 3 — classifying stories; Phase 1 — looking up company aliases & tech terms        |
| `output_templates.md` | Phase 5 — rendering final output; selecting template by user request                    |
| `workflow.md`         | All Phases — orchestrating the end-to-end workflow; time budgeting; monthly maintenance |

### File Interconnection Map

```
┌─────────────────┐      ┌────────────────────┐     ┌───────────────┐     ┌──────────────────┐
│  search_queries │────▶ │  news_sources      │────▶│  Classify &   │────▶│ output_templates │
│  (discover)     │      │  (browse & verify) │     │  Prioritize   │     │   (generate)     │
└─────────────────┘      └────────────────────┘     └───────────────┘     └──────────────────┘
                                    ▲                        ▲
                                    │                        │
                                    └────── taxonomy.md ─────┘
                                         (shared vocabulary)
```

---

## Execution Workflow

### Phase 0: Determine Briefing Type & Time Scope

**Before any tool calls**, ask the user (if not already clear):

1. **Briefing Type**: Daily / Weekly / Monthly / Custom Topic?
2. **Time Scope**: Last 24 hours / Last 7 days / Last 30 days / Custom date range?
3. **Output Format**: Standard / Brief / Thread / Markdown Report / Presentation / Custom?
4. **Focus Area** (optional): All categories / Specific category (e.g., only hardware, only China ecosystem)?

**Default if user doesn't specify**:

- Type: Daily
- Scope: Last 24 hours
- Format: Standard
- Focus: All categories

**Map to workflow.md**:

- Daily → `workflow.md` Section "Daily Workflow"
- Weekly → `workflow.md` Section "Weekly Workflow"
- Monthly → `workflow.md` Section "Monthly Workflow"

---

### Phase 1: Information Gathering

Consult `workflow.md` for the appropriate recipe, then execute the corresponding steps from `search_queries.md` and `news_sources.md`.

#### Step 1.1: Execute Search Queries

**Tool**: `WebSearch` (or equivalent web search tool)

**Source**: `search_queries.md` → Select the appropriate recipe:

- Daily Briefing → Recipe A (5 queries)
- Weekly Roundup → Recipe B (8 queries)
- Monthly Deep Dive → Recipe C (12 queries)
- Custom Topic → Recipe D + user-specified filters

**Parameters**:

- `return_format`: markdown
- `with_images_summary`: false
- `timeout`: 20 seconds per source
- Fetch only from publicly accessible sources listed in `news_sources.md`

**Output**: A list of 20–50 URLs with headlines and snippets.

---

#### Step 1.2: Fetch Tier 1 Sources Directly

**Tool**: `mcp__web_reader__webReader`

**Source**: `news_sources.md` → Tier 1 section

Directly fetch the homepage or RSS feed of:

- The Robot Report
- IEEE Spectrum — Robotics
- TechCrunch — Robotics
- Robotics Business Review
- (Add others based on briefing type)

**Parameters**:

- `url`: [homepage URL from news_sources.md]
- `return_format`: markdown
- `with_images_summary`: false
- Process only URLs from verified sources in `news_sources.md`

**Output**: Recent headlines (last 24h / 7d / 30d based on scope).

---

#### Step 1.3: Fetch arXiv Papers

**Tool**: `mcp__arxiv__readURL` (if available) or `WebSearch` with arXiv-specific queries

**Source**: `search_queries.md` → Section "6. Academic Research (arXiv)"

Execute 2–3 arXiv queries:

```
cat:cs.RO AND ("embodied AI" OR "robot learning" OR "VLA") submittedDate:[today - 7d TO today]
```

**Output**: 5–10 recent papers with abstracts.

---

#### Step 1.4: Fetch Company Blogs & Official Announcements

**Tool**: `mcp__web_reader__webReader`

**Source**: `news_sources.md` → Tier 2 (Company Blogs) + Tier 4 (China Ecosystem)

Fetch from:

- Figure AI Blog
- Physical Intelligence Blog
- Tesla AI Blog
- Unitree Blog (Chinese + English)
- AGIBOT WeChat Official Account (if accessible)
- (Add others based on focus area)

**Fetch constraints**:

- Only process URLs from search results and sources listed in `news_sources.md`
- Skip content requiring authentication
- Timeout: 15 seconds per URL

**Output**: Recent announcements (last 7d / 30d based on scope).

---

### Phase 2: Content Extraction & Deduplication

For each fetched URL:

1. **Extract**:
   - Headline
   - Publication date
   - Source name
   - Summary (first 2–3 paragraphs or abstract)
   - Key entities: companies, models, hardware platforms (use `taxonomy.md` for reference)

2. **Deduplicate**:
   - If multiple sources cover the same story, keep the one with the most detail
   - Merge information if they provide complementary details

3. **Discard**:
   - Stories older than the time scope
   - Irrelevant content (use `search_queries.md` Section 1.4 "Noise Exclusion Filter")
   - Duplicate announcements

**Output**: A deduplicated list of 15–30 stories with extracted metadata.

---

### Phase 3: Classification & Prioritization

Consult `taxonomy.md` to classify each story.

#### Step 3.1: Assign Primary Category

Use `taxonomy.md` → Section "